大數(shù)據(jù)技術(shù)之Flume

1、 大數(shù)據(jù)技術(shù)之標(biāo)題一、Flume簡介1) Flume提供一個分布式的，可靠的，對大數(shù)據(jù)量的日志進(jìn)行高效收集、聚集、移動的服務(wù)，F(xiàn)lume只能在Unix環(huán)境下運行。2) Flume基于流式架構(gòu)，容錯性強，也很靈活簡單。3) Flume、Kafka用來實時進(jìn)行數(shù)據(jù)收集，Spark、Storm用來實時處理數(shù)據(jù)，impala用來實時查詢。二、Flume角色2.1、Source用于采集數(shù)據(jù)，Source是產(chǎn)生數(shù)據(jù)流的地方，同時Source會將產(chǎn)生的數(shù)據(jù)流傳輸?shù)紺hannel，這個有點類似于Java IO部分的Channel。2.2、Channel用于橋接Sources和Sinks，類似于一個隊列。2.3

2、、Sink從Channel收集數(shù)據(jù)，將數(shù)據(jù)寫到目標(biāo)源(可以是下一個Source，也可以是HDFS或者HBase)。2.4、Event傳輸單元，F(xiàn)lume數(shù)據(jù)傳輸?shù)幕締卧?，以事件的形式將?shù)據(jù)從源頭送至目的地。三、Flume傳輸過程source監(jiān)控某個文件或數(shù)據(jù)流，數(shù)據(jù)源產(chǎn)生新的數(shù)據(jù)，拿到該數(shù)據(jù)后，將數(shù)據(jù)封裝在一個Event中，并put到channel后commit提交，channel隊列先進(jìn)先出，sink去channel隊列中拉取數(shù)據(jù)，然后寫入到hdfs或者HBase中。四、Flume部署及使用4.1、文件配置flume-env.sh涉及修改項：JAVA_HOME=/home/admin/mo

3、dules/jdk1.8.0_1214.2、案例4.2.1、案例一目標(biāo)：Flume監(jiān)控一端Console，另一端Console發(fā)送消息，使被監(jiān)控端實時顯示。分步實現(xiàn)：1) 創(chuàng)建Flume Agent配置文件flume-telnet.conf# Name the components on this agenta1.sources = r1a1.sinks = k1a1.channels = c1# Describe/configure the sourcea1.sources.r1.type = netcata1.sources.r1.bind = localhosta1.sources.r1

4、.port = 44444# Describe the sinka1.sinks.k1.type = logger# Use a channel which buffers events in memorya1.channels.c1.type = memorya1.channels.c1.capacity = 1000a1.channels.c1.transactionCapacity = 100# Bind the source and sink to the channela1.sources.r1.channels = c1a1.sinks.k1.channel = c12) 安裝te

5、lnet工具$ sudo rpm -ivh telnet-server-0.17-59.el7.x86_64.rpm $ sudo rpm -ivh telnet-0.17-59.el7.x86_64.rpm3) 判斷44444端口是否被占用$ netstat -an | grep 444444) 先開啟flume先聽端口$ bin/flume-ng agent -conf conf/ -name a1 -conf-file conf/flume-telnet.conf -Dflume.root.logger=INFO,console5) 使用telnet工具向本機的44444端口發(fā)送內(nèi)容$

6、telnet localhost 444444.2.2、案例二目標(biāo)：實時監(jiān)控hive日志，并上傳到HDFS中分步實現(xiàn)：1) 拷貝Hadoop相關(guān)jar到Flume的lib目錄下（要學(xué)會根據(jù)自己的目錄和版本查找jar包）$ cp share/hadoop/common/lib/hadoop-auth-2.5.0-cdh5.3.6.jar ./lib/$ cp share/hadoop/common/lib/commons-configuration-1.6.jar ./lib/$ cp share/hadoop/mapreduce1/lib/hadoop-hdfs-2.5.0-cdh5.3.6.

7、jar ./lib/$ cp share/hadoop/common/hadoop-common-2.5.0-cdh5.3.6.jar ./lib/$ cp ./share/hadoop/hdfs/lib/htrace-core-3.1.0-incubating.jar ./lib/$ cp ./share/hadoop/hdfs/lib/commons-io-2.4.jar ./lib/尖叫提示：標(biāo)紅的jar為1.99版本flume必須引用的jar2) 創(chuàng)建flume-hdfs.conf文件# Name the components on this agenta2.sources = r2a

8、2.sinks = k2a2.channels = c2# Describe/configure the sourcea2.sources.r2.type = execmand = tail -F /home/admin/modules/apache-flume-1.7.0-bin/my_custom_logs.txta2.sources.r2.shell = /bin/bash -c# Describe the sinka2.sinks.k2.type = hdfsa2.sinks.k2.hdfs.path = hdfs:/linux01:8020/flume/%Y%m%d/%H#上傳文件的

9、前綴a2.sinks.k2.hdfs.filePrefix = logs-#是否按照時間滾動文件夾a2.sinks.k2.hdfs.round = true#多少時間單位創(chuàng)建一個新的文件夾a2.sinks.k2.hdfs.roundValue = 1#重新定義時間單位a2.sinks.k2.hdfs.roundUnit = hour#是否使用本地時間戳a2.sinks.k2.hdfs.useLocalTimeStamp = true#積攢多少個Event才flush到HDFS一次a2.sinks.k2.hdfs.batchSize = 1000#設(shè)置文件類型，可支持壓縮a2.sinks.k2.

10、hdfs.fileType = DataStream#多久生成一個新的文件a2.sinks.k2.hdfs.rollInterval = 600#設(shè)置每個文件的滾動大小a2.sinks.k2.hdfs.rollSize = 134217700#文件的滾動與Event數(shù)量無關(guān)a2.sinks.k2.hdfs.rollCount = 0#最小冗余數(shù)a2.sinks.k2.hdfs.minBlockReplicas = 1# Use a channel which buffers events in memorya2.channels.c2.type = memorya2.channels.c2.c

11、apacity = 1000a2.channels.c2.transactionCapacity = 100# Bind the source and sink to the channela2.sources.r2.channels = c2a2.sinks.k2.channel = c23) 執(zhí)行監(jiān)控配置$ bin/flume-ng agent -conf conf/ -name a2 -conf-file conf/flume-hdfs.conf4.2.3、案例三目標(biāo)：使用flume監(jiān)聽整個目錄的文件分步實現(xiàn)：1) 創(chuàng)建配置文件flume-dir.confa3.sources = r3a

12、3.sinks = k3a3.channels = c3# Describe/configure the sourcea3.sources.r3.type = spooldira3.sources.r3.spoolDir = /home/admin/modules/apache-flume-1.7.0-bin/uploada3.sources.r3.fileHeader = true#忽略所有以.tmp結(jié)尾的文件，不上傳a3.sources.r3.ignorePattern = ( *.tmp)# Describe the sinka3.sinks.k3.type = hdfsa3.sinks

13、.k3.hdfs.path = hdfs:/linux01:8020/flume/upload/%Y%m%d/%H#上傳文件的前綴a3.sinks.k3.hdfs.filePrefix = upload-#是否按照時間滾動文件夾a3.sinks.k3.hdfs.round = true#多少時間單位創(chuàng)建一個新的文件夾a3.sinks.k3.hdfs.roundValue = 1#重新定義時間單位a3.sinks.k3.hdfs.roundUnit = hour#是否使用本地時間戳a3.sinks.k3.hdfs.useLocalTimeStamp = true#積攢多少個Event才flush

14、到HDFS一次a3.sinks.k3.hdfs.batchSize = 1000#設(shè)置文件類型，可支持壓縮a3.sinks.k3.hdfs.fileType = DataStream#多久生成一個新的文件a3.sinks.k3.hdfs.rollInterval = 600#設(shè)置每個文件的滾動大小a3.sinks.k3.hdfs.rollSize = 134217700#文件的滾動與Event數(shù)量無關(guān)a3.sinks.k3.hdfs.rollCount = 0#最小冗余數(shù)a3.sinks.k3.hdfs.minBlockReplicas = 1# Use a channel which buffers events in memorya3.channels.c3.type = memorya3.channels.c3.capacity = 1000a3.channels.c3.transactionCapacity = 100# Bind the source and sink to the channela3.sources.r3.channels = c3a3.sinks.k3.cha